跳至主要內容
Kyeo AI
AI 圖片模型
Tongyi-MAI
Z-Image

Z-Image

Kyeo AI 目前的 Z-Image 是一項操作精簡的文字生成圖片服務：公開資料服務原始價格為每張圖 0.8 credits，本地整數點數錢包向上取整為 1 點數；提示詞上限為 1,000 個字元，可選擇 5 種長寬比。這個流程不接受圖片輸入，也不提供編輯功能。

使用此模型查看點數費用回到模型廣場
能力標籤
文字生成圖片
點數消耗
1 點數 / 次
提示詞上限
1,000 個字元
上傳限制
不適用（不接受圖片）
30 秒快速瞭解
  • 目前只支援文字生成圖片，不接受圖片輸入或提供編輯功能
  • 介面提供提示詞與 `aspect_ratio`，可選擇 5 種常用長寬比
  • 提示詞上限為 1,000 個字元，適合結構清楚、重點明確的描述
  • 公開資料服務原始價格為每張圖 0.8 credits；Kyeo 對總價向上取整，單圖為 1 點數

熱門搜尋: Z-Image 怎麼用 · Z-Image 價格 · Z-Image 1 點數

示例

Visual concept

Visual concept

Aspect Ratio: 9:16 Vertical Duration: 15 Seconds Style: Premium street-fashion commercial, cinematic lifestyle advertising, realistic camera physics, luxury tech campaign, viral social-media edit. IMPORTANT: Treat each storyboard panel as an individual full-screen shot. Read left-to-right, top-to-bottom. Do NOT animate the collage. Every frame becomes a separate cinematic scene. Use the exact white-and-green wireless headphones from the reference images. Preserve ear-cup shape, hinge design, logo placement, premium matte finish, soft leather cushions, realistic materials, and proportions throughout all shots. ### Shot 01 (0.0–1.2s) Wide cinematic establishing shot. Young woman in oversized red shirt and black outfit walks through a red-brick urban alley. Headphones hang casually from her hand. Slow handheld push forward. Warm late-afternoon light. Fashion-commercial atmosphere. ### Shot 02 (1.2–2.2s) Macro product detail. Hand raises the white-and-green headphones. Shallow depth of field. Premium texture visible. Soft city bokeh. Gentle camera drift. ### Shot 03 (2.2–3.3s) Medium shot. She places headphones over her ears. Hair moves slightly in the wind. Match cut synchronized to soundtrack beat. Smooth handheld realism. ### Shot 04 (3.3–4.4s) Wide tracking shot. She walks deeper into the alley. Camera follows beside her. Natural parallax from buildings. Urban confidence. ### Shot 05 (4.4–5.5s) Extreme close-up of ear cup. Green accents catch reflected light. Premium material detail. Cinematic rack focus. ### Shot 06 (5.5–6.7s) Medium close-up profile. Eyes closed. Calm expression. Noise-canceling serenity. Slow push-in. Background softly blurred. ### Shot 07 (6.7–7.9s) Low-angle wide shot crossing a city intersection. Golden-hour reflections. Confident stride. Headphones centered naturally within frame. ### Shot 08 (7.9–9.1s) Tight side profile. Hair texture and skin detail contrasted against smooth headphone surface. Passing cars create subtle motion streaks. ### Shot 09 (9.1–10.3s) Medium portrait. She lifts one ear cup slightly. Small smile. Connecting with the city atmosphere. Gentle camera sway. ### Shot 10 (10.3–11.5s) Wide follow shot from behind. Walking toward a graffiti-covered wall. Red-and-black environment complements the green-and-white product. ### Shot 11 (11.5–12.8s) Macro hero product shot. Headphones rotate slowly in her hand. Dynamic reflections reveal premium craftsmanship. Sharp logo visibility. ### Shot 12 (12.8–15.0s) Hero ending. Woman seated on urban staircase. Headphones worn confidently. Soft backlight creates halo around hair. Slight breeze. Slow cinematic push-in. Peaceful yet powerful expression. Product remains crisp, dominant, and aspirational. **Camera Language** * Realistic handheld movement * Slow push-ins * Natural parallax * Motivated pans and tilts * Selective focus transitions * High-end commercial cinematography * No floating drone moves * No artificial zoom effects * No CGI camera tricks **Transitions** * Hard cuts on beat * Match cuts using hand movement * Match cuts using head turns * Momentum-driven edits * No wipes * No dissolves * No graphic overlays **Sound Design** * Distant city traffic * Soft wind * Footsteps on pavement * Clothing movement * Headphone handling sounds * Subtle headphone power-click effects * Deep ambient bass pulse * Modern cinematic electronic beat * Emotional luxury-tech atmosphere **Final Frame** Hold for 0.5 seconds on the seated hero shot. White-and-green headphones remain perfectly visible, logo clear, premium lighting, cinematic depth, luxury advertising finish, viral social-media campaign quality, ultra-realistic 4K HDR.

Visual concept

Visual concept

Ultra-realistic luxury product photography of a white creamy glass reed diffuser bottle labeled “WHITE OUDH – Private Collection – RITUALS” in elegant gold serif typography. The bottle has a curved rectangular shape with soft edges, a glossy reflective surface, and a gold wax-seal style emblem centered near the top. Thin black diffuser reeds extend upward from the bottle neck. The bottle is covered in realistic water AND ICE droplets and condensation, as if freshly misted after rain. It is placed among deep black roses and dark burgundy petals with visible water droplets, surrounded by rich green leaves. The scene is moody and dramatic with a dark botanical background, softly blurred (shallow depth of field, bokeh effect). Lighting is cinematic and low-key, with soft diffused highlights reflecting off the wet glass surface. Cool-toned lighting (deep greens and midnight blues) enhances the dark, luxurious atmosphere. The gold emblem and lettering subtly catch the light, creating a premium, high-end aesthetic. Composition: vertical portrait orientation (4:5 ratio), centered product placement, slightly angled bottle resting naturally among roses. Foreground roses partially frame the bottom of the bottle. Background is dark foliage fading into shadow. Texture emphasis: hyper-detailed water droplets, glossy glass reflections, velvety rose petals, matte black reeds. Style: high-end commercial fragrance advertising, macro lens, 85mm, f/2.0, shallow depth of field, photorealistic, 8K resolution, sharp focus on bottle, dramatic contrast, rich blacks, luxury perfume campaign aesthetic.

Visual concept

Visual concept

[BRAND NAME] + [COLOR] Act as a Senior CGI Artist and Brand Identity Director specializing in premium logo materializations. Your reference aesthetic: a logo mark that appears to have been pushed outward from behind a metallic or matte surface — like a relief stamp pressed from the reverse side of a metal sheet, a hallmark embossed on luxury packaging, or a raised seal on official documents. The logo does not exist as a separate standing object placed on a surface. It exists as a raised relief, a bulge, an outward protrusion that is part of the surface itself. BRAND INTELLIGENCE SYSTEM Before executing any phase, resolve these parameters: (1) LOGO GEOMETRY — identify [BRAND NAME]'s primary icon mark in its simplest most reduced form — the mark that works at any scale, (2) COLOR PALETTE — based on [COLOR], build the full tonal system from the lightest specular highlight to the deepest shadow tone, determine if the overall mood reads warm, cool, or neutral, (3) SURFACE MATERIAL — based on [COLOR] determine the surface character: silver or grey becomes brushed or circular-grain steel, gold or champagne becomes warm brushed metal with radial grain, black becomes anodized aluminum or matte carbon fiber, white becomes matte ceramic or chalk plaster, any saturated hue becomes anodized colored metal holding the hue in lit areas and desaturating toward black in deepest shadows, (4) LIGHT DIRECTION — determine the most flattering single light angle to reveal the emboss topography — where highlights land on the raised ridge peaks and where shadows fall on the descending walls. PHASE 1: SURFACE & ATMOSPHERE The entire image is one single continuous material surface — not a background with an object placed on it, but one unified physical plane filling the entire canvas. This surface is made of the SURFACE MATERIAL resolved above. The surface has a subtle radial gradient in its lighting — brighter near the center where the embossed logo sits, gradually darkening toward all four edges as the light falls off naturally. If the material is brushed metal, the grain direction is radial or concentric — circular polishing marks emanating from the center, catching the light differently at each angle. Apply fine uniform grain texture across the entire surface at ISO 800 equivalent — this gives tactility, depth, and prevents any digital flatness. The surface must feel physical and touchable — like a metal plate, a luxury tin lid, or a premium embossed card stock you could run your finger across and feel the texture. PHASE 2: THE EMBOSS — CRITICAL The [BRAND NAME] logo mark is rendered as a full bas-relief — think of a coin, a commemorative medal, or a luxury brand stamp where the ENTIRE logo shape rises as one unified solid mass from the surface. Not just the outline. Not just the stroke path. The complete filled silhouette of the logo pushes outward from the surface as a single continuous raised form — like the head on a coin pressed from a die. The interior of the logo is NOT hollow or open. The entire logo shape — every filled area, every curve, every solid region — rises together as one piece, at a uniform height above the surrounding flat surface. The top face of this raised form is slightly convex — gently domed, not perfectly flat — so it catches the key light across its full width and creates a smooth highlight gradient from the lit side to the shadow side. The transition from the flat surrounding surface up to the raised logo form follows a smooth beveled wall — curved, not sharp — like the edge of a coin. This beveled wall catches the key light on the lit side and falls into shadow on the opposite side, creating the dimensional separation between the raised form and the flat background. Any negative spaces WITHIN the logo — such as the bite in the Apple, the interior counters of letterforms — are recessed BACK to the level of the flat surface or slightly below, creating clean negative voids within the overall raised form. These recessed areas have their own small beveled walls descending inward, catching light in reverse. The overall raised height must feel substantial — like a thick coin or a heavy embossed seal — not a shallow bump. The logo geometry must be completely faithful to [BRAND NAME]'s actual mark in correct proportions and correct silhouette. PHASE 3: LIGHTING One primary soft area light from upper-left at approximately 10 to 11 o'clock position — broad and diffused, wrapping around the curved walls of the raised ridge to create the characteristic highlight peak at the top and gradual shadow descent on the far wall. Color temperature is determined by [COLOR]: cool white light for silver, grey, and blue tones — warm amber light for gold, copper, and earth tones — neutral daylight for white and black. One very subtle fill light from lower-right at 10 to 15 percent intensity of the key — just enough to retain material detail in the deepest shadow areas of the ridge walls without lifting them enough to flatten the form. No hard shadows anywhere on the surface. No spotlight pools. No rim light. The lighting exists entirely to reveal the three-dimensional topography of the emboss — every curve, every height transition, every subtle variation in the surface. PHASE 4: TYPOGRAPHY In the lower portion of the canvas, below the embossed logo, on the same continuous surface — a minimal typographic lockup. The typography is either very subtly embossed in the same material as the surface using the same raised-ridge technique at much smaller scale, or rendered as clean flat text in the lightest tone of [COLOR] sitting just at the surface level. Element 1: [BRAND NAME]'s flat logo mark at small scale — approximately 3 to 4 percent of canvas width — as a simple flat mark, not embossed. Element 2: [BRAND NAME] wordmark in clean geometric sans-serif spaced capitals directly below the icon mark. Element 3: one short italic or light-weight descriptor line significantly smaller than the wordmark — autonomously generate a relevant subtitle for [BRAND NAME] such as a founding year, a product category, a brand division, or a core tagline. All three elements stacked vertically, centered, with generous spacing. No decorative elements. No lines or rules. Pure typographic restraint. TECH SPECS The single most important instruction in this entire prompt: the logo must read as PUSHED OUTWARD FROM the surface, rising toward the viewer, proud and raised. If the result looks like a 3D logo object floating above a background it is wrong. If the result looks like a depression or hole cut into the surface it is wrong. The correct result looks like you are holding a sheet of metal that someone pressed a die stamp against from the back — and you are looking at the front where the relief has formed. Ray Tracing enabled for accurate self-shadowing — the raised ridge must cast a real shadow onto the flat surrounding surface. No Depth of Field — the entire surface is in perfect sharp focus from edge to edge. Tone mapping: preserve highlight detail — the brightest specular on the ridge peak should be brilliant and tight but never blown out or featureless. Film grain applied uniformly across the full final render. No chromatic aberration. No lens distortion. No added vignette beyond what the surface lighting gradient naturally creates. Anti-aliasing maximum — the ridge edges must be perfectly clean. Mood: a maker's hallmark, a foundry seal, a luxury embossed cover — the quiet confidence of a brand that lets its materiality speak.

Visual concept

Visual concept

Dreamy sunlit portrait of a young woman with long soft light-brown hair, delicate facial features, fair luminous skin, natural rosy lips, and expressive hazel-green eyes. Wearing an oversized cream-white knitted sweater, sitting beside a large window with warm golden morning sunlight streaming through. Soft shadows from the window frame fall across her face and clothing, creating a cozy and intimate atmosphere. Gentle wind slightly tousling her hair, relaxed pose with one hand resting near her face, minimal makeup, natural beauty, Scandinavian-inspired aesthetic, soft pastel tones, shallow depth of field, creamy bokeh, cinematic lighting, ultra-realistic skin texture, editorial fashion photography, emotional storytelling, serene and elegant mood, photorealistic, masterpiece quality, 8K HDR, 85mm lens, f/1.4. Enhanced Cinematic Version: Golden-hour indoor portrait, beautiful young woman bathed in warm sunlight, sitting peacefully near a sunlit window, long flowing light-brown hair glowing with backlight, oversized cozy knit sweater, soft freckles, radiant skin, dreamy gaze toward the camera, delicate facial symmetry, warm ivory and beige tones, luxury lifestyle photography, airy atmosphere, natural light masterpiece, cinematic depth, soft highlights, premium fashion editorial, Vogue-style portrait, ultra-detailed eyes, realistic textures, high-end photography, shallow depth of field, warm glow, photorealistic, 8K. Negative Prompt: low quality, blurry, overexposed highlights, underexposed shadows, bad anatomy, distorted face, extra fingers, extra limbs, asymmetrical eyes, plastic skin, heavy makeup, oversaturated colors, noise, grain, watermark, text, logo, motion blur, CGI, cartoon, anime, unrealistic proportions, low resolution, artifacts. Style Keywords: golden hour, cozy aesthetic, soft sunlight, dreamy portrait, warm glow, cream sweater, editorial photography, Scandinavian minimalism, airy atmosphere, natural beauty, cinematic realism, luxury lifestyle

Visual concept

Visual concept

{ "model": "gemini-nano-banana-pro", "mode": "image", "aspect_ratio": "4:5", "quality": "ultra", "style": "cinematic_realism", "scene": { "setting": "busy neon-lit city street at night, wet road after rain, glowing signboards and traffic behind subject", "weather": "light drizzle just ended, reflective pavement, humid atmosphere", "background": "soft bokeh city lights, moving cars, faint pedestrian motion blur, tall buildings with colorful signage", "time": "night" }, "subject": { "type": "young woman", "pose": "standing in the middle of a crosswalk, holding a smartphone close to camera", "expression": "natural warm smile, candid and joyful", "hair": "slightly messy hair blown across face by wind", "wardrobe": "dark coat or blazer, minimal accessories", "props": ["smartphone"] }, "composition": { "framing": "close-up portrait with wide-angle distortion, subject centered, phone in foreground", "camera_distance": "very close, intimate street-photo feel", "perspective": "wide lens perspective, slight fisheye look, background stretching outward", "depth_of_field": "shallow to medium, subject readable but background lights creamy", "foreground_detail": "phone and hands slightly soft due to motion and lens character" }, "camera": { "lens": "24mm wide-angle (slight fisheye character)", "aperture": "f/1.8", "shutter_speed": "1/40", "iso": "1600", "focus": "face focus, slight falloff on phone" }, "lighting": { "key_light": "neon signage spill lighting face with mixed cyan/green and warm highlights", "fill": "streetlight bounce from wet ground", "mood": "dreamy urban nightlife, soft contrast, romantic realism" }, "color_grading": { "palette": "teal, cyan, green neon mixed with warm skin tones", "contrast": "soft contrast with lifted shadows", "highlights": "glowy bloom on neon signs", "film_emulation": "35mm film look" }, "texture_and_detail": { "film_grain": "strong visible grain, authentic", "bloom": "subtle halation on lights", "sharpness": "slightly imperfect, organic street photography", "skin_texture": "natural pores, no plastic smoothing" }, "motion": { "hair_movement": "wind-blown strands across face", "background_motion": "light streaking from cars and walking people", "camera_feel": "handheld candid shot" }, "negative_prompt": [ "plastic skin", "overly sharp face", "clean studio lighting", "anime", "cartoon", "hyper-symmetrical face", "over-saturated neon", "HDR look", "fake depth blur edges", "extra fingers", "distorted hands", "text artifacts", "logo distortions", "AI glossy skin", "perfectly clean background" ] }

Visual concept

Visual concept

Create a Cannes-level sports fashion advertising poster for an original performance sneaker brand named "AEROSTEP". Use ultra-realistic live-action commercial photography. Preserve the exact structural logic of the reference image: an extreme low-angle street-level shot on a sunny boardwalk or beachside promenade, one adult athletic woman dominating the frame through forced perspective as she bends down to tie her sneaker, with clear blue sky filling most of the background, palm trees and a very small number of casual pedestrians receding behind her, and the shoe action as the absolute visual hero. Style lock: premium sportswear campaign photography, street-fashion realism, ultra-low-angle lifestyle advertising, bright midday natural sunlight, documentary-meets-editorial commercial image, no illustration, no CGI look, no style drift. Port logic: make the brand super-symbol the moment of bending down to tie the sneaker from an exaggerated ground-up perspective. The identity of the campaign must be the oversized near-camera shoe, the athletic fold of the body, and the giant-on-the-boardwalk scale contrast. This tying gesture must feel instantly ownable, dynamic, and recognizable as the signature visual code of the brand. Transit logic: reduce all non-essential environmental information so the image reads more like a flagship sneaker poster. Keep only the minimum boardwalk context needed to support scale and realism: a cleaner horizon line, fewer pedestrians, softer storefront detail, and more open negative space in the sky. The final image must feel more controlled, more premium, and more product-led. Core composition: full-body female figure seen from a dramatic ground-up perspective, body bent forward in a compact athletic fold, one hand tying the sneaker lace near the camera, one leg strongly foreshortened and dominant in the foreground, the torso and head leaning diagonally downward, short blonde hair moving lightly in the breeze. The shoe closest to camera must be the main product focal point. Keep the open promenade street, a very small number of distant figures, and minimal palm-tree cues only as support. Product focus: make the hero product a premium silver-white performance sneaker with technical mesh, layered panel construction, subtle reflective accents, sculpted outsole, and realistic lace tension. The shoe must feel contemporary, stylish, and performance-oriented. Preserve the same tying-shoelace action and the same near-camera product emphasis. The sneaker remains the absolute visual hero. Wardrobe styling: an adult woman wearing a minimal retro-sport knit set in muted sage green and deep green, consisting of a fitted ribbed crop top and matching athletic briefs or sport bottoms, paired with scrunched white socks. Styling should feel clean, sunlit, youthful, and premium, with no extra accessories beyond the sportswear and shoes. Human realism: adult female only, natural anatomy, realistic athletic body proportions, believable perspective distortion caused only by the extreme low-angle lens, correct shoulders, arms, pelvis, thighs, knees, calves, ankles, feet, and hands. Real skin only with subtle pores, natural skin texture, realistic limb compression, soft body volume, slight natural imperfections, no mannequin skin, no plastic smoothing, no anatomical errors. Lighting: strong natural midday sunlight with clear directional shadows on the pavement, crisp highlight on skin and shoes, luminous blue-sky ambience, realistic street contrast, and no muddy blacks. Keep the scene hot, bright, and sharp, with clean shadow edges and premium commercial clarity. Environment: a bright coastal urban promenade with pale pavement, one or two palm trees, a few very distant pedestrians, and extremely simplified low storefront forms far in the background. Preserve the same sense of scale contrast between the giant near-camera athlete and the tiny background world, but keep the environment quieter and more minimal than the reference. No extra props, no vehicles, no visual clutter. Color strategy: dominant palette of sky blue, sunlit skin tones, silver-white sneaker highlights, muted sage green, deep green, and warm concrete neutrals. Keep the image clean, summery, sporty, and modern. Art direction goals: product-first hierarchy, brand super-symbol clarity, flagship restraint, bold forced-perspective impact, premium sneaker desirability, authentic street-sport energy, clean commercial finish, no text, no watermark, no logo clutter, no black blotches, no low-resolution details. Negative prompt: different camera angle, eye-level shot, missing low-angle perspective, distorted anatomy, oversized deformed limbs, extra fingers, broken hands, warped sneaker shape, unreadable laces, muddy pavement, cloudy sky, busy street, excessive crowd, distracting storefront detail, plastic skin, CGI look, illustration style, watermark, headline text, logo clutter, black blotches, low-resolution shoe details.

一句話瞭解

這個模型表現如何？

如果任務是直接從文字生成新圖片，Z-Image 提供清楚且容易估算成本的選擇。需要修改現有圖片、多張參考圖或解析度等級時，可再比較對應模型。

它擅長什麼
Z-Image 在 Kyeo 的定位是操作精簡的文字生成圖片選項。目前介面不提供圖片輸入、編輯、參考圖或畫質等級；是否適合任務，應依所需輸入與控制項判斷，而不是預先把結果視為草稿或成品。
適合哪些使用者
適合從文字製作封面、社群圖片、海報概念與視覺風格，也適合想用固定成本嘗試多種提示詞的個人或團隊。
為什麼要在 Kyeo AI 使用
Kyeo 將目前的使用範圍清楚限定為提示詞加上 `aspect_ratio`：提示詞最長 1,000 個字元，長寬比有 5 種，每次 1 點數。使用者可在生成前直接確認輸入限制與成本。

關鍵參數總覽

協助你快速評估此模型是否符合你的業務情境。

模型類別
AI 圖片模型
廠商
Tongyi-MAI
模型家族
Z-Image
串接方式
目前站內串接
平台模型識別碼
z-image
執行方式
非同步任務
提示詞上限
1,000 個字元
上傳限制
不適用（不接受圖片）

其他名稱

同一個模型在官方文件、社群討論與 API 串接層可能有不同稱呼，這裡幫你統一整理。

z-image
Z Image
ZImage

使用者常見問題

從實際任務、輸入條件和結果要求出發，整理選擇模型時最需要確認的問題。

Z-Image 怎麼用
Z-Image 價格
Z-Image 1 點數
Z-Image 提示詞上限
Z-Image 支援圖生圖嗎
Z-Image 支援哪些比例
Z-Image 適合哪些任務
Z-Image 與 Nano Banana 的差異
Z-Image 與 GPT Image 1.5 的差異
Z-Image 能否編輯上傳的圖片

選擇指南

還在猶豫該用哪個模型嗎？先查看這幾個關鍵決策點。

1
只需要從文字生成新圖片時

如果任務不需要上傳現有圖片，也不依賴參考圖、編輯或畫質等級，Z-Image 的輸入項目少、單次成本固定，選擇過程相對直接。

2
需要編輯或參考圖時

如果任務必須修改現有圖片、保留參考圖中的主體，或選擇解析度與畫質等級，應先比較 Nano Banana、GPT Image 1.5 或 Flux 2 Flex。

3
先用明確的提示詞測試

先決定目標比例，再用一則包含主體、場景、構圖與風格的提示詞生成圖片。依實際結果繼續調整；只有需要目前介面沒有的控制項時，才需要更換模型。

使用者常搜尋的比較

把常見候選放在同一任務下比較，協助你確認輸入方式、控制項和成本差異。

Z-Image 與 Nano Banana 在圖片輸入與價格上有何不同？
只做文字生成圖片時，Z-Image 與 Nano Banana 該怎麼選？
Z-Image 與 GPT Image 1.5 的編輯能力有何不同？
何時應從 Z-Image 改用 GPT Image 1.5？
Z-Image 與 Flux 2 Flex 的解析度與價格有何不同？
哪些任務需要 Flux 2 Flex，而不是 Z-Image？

模型比較表

還在猶豫該選哪一個嗎？這張表協助你快速看清目前模型與替代方案的核心差異。

目前站內定位
Z-Image
低成本文字生成圖片
Nano Banana
同時支援文字生成圖片與圖片編輯
GPT Image 1.5
支援文字生成圖片、圖片編輯與畫質等級
工作流程邊界
Z-Image
僅支援文字生成圖片
Nano Banana
文字生成圖片 + 圖片編輯
Flux 2 Flex
文字生成圖片 + 圖片編輯
提示詞上限
Z-Image
1,000 個字元
Nano Banana
5,000 個字元
GPT Image 1.5
未公布字元上限
目前價格口徑
Z-Image
固定 1 點數 / 次
Nano Banana
固定 3 點數 / 次
Flux 2 Flex
14–24 點數 / 次
更適合的階段
Z-Image
只需從文字生成新圖片
GPT Image 1.5
需要圖片編輯或畫質等級
Nano Banana
需要在同一個流程中生成或修改圖片

實際使用體驗

結合公開資料、目前站內邊界和典型任務，說明模型適合與不適合的情況。

控制項少不代表輸出畫質低

Kyeo 目前只開放提示詞與長寬比，這表示介面範圍精簡，不能據此判定圖片只能當作草稿。結果是否符合需求，仍應依具體提示詞與實際生成圖片判斷。

價格與工作流程要分開判斷

固定 1 點數降低了單次嘗試成本，但價格不能取代功能判斷。先確認是否只需要文字生成圖片，再依實際結果決定繼續調整，或改用具備編輯、參考圖或畫質等級的模型。

功能詳解

純文字生圖入口

只接收文字提示詞，不接收參考圖；適合從零生成新圖片，不適合局部修改現有素材。

5 種長寬比

目前可選擇 1:1、4:3、3:4、16:9 與 9:16，涵蓋正方形、橫向與直向構圖。介面不另設像素尺寸或畫質等級。

明確的提示詞與計費限制

提示詞最長 1,000 個字元，每次生成固定消耗 1 點數。目前不會依長寬比、畫質或解析度套用其他價格等級。

適合做什麼

概念草圖

透過文字快速確認主體、場景、構圖與風格是否符合預期，再依實際結果調整提示詞。

封面與社群圖片

可用 1:1、16:9 或 9:16 等比例生成文章封面、短影片封面與社群首圖；是否直接使用，應以實際輸出結果為準。

提示詞與構圖探索

固定的單次成本與少量控制項適合逐次比較不同主題、構圖或風格描述，但目前頁面不提供批次生成功能。

提示詞技巧

依重要性安排提示詞

在 1,000 個字元內，優先寫清主體、場景、構圖、光線與風格，把必要條件放在前面，避免堆疊互相衝突的修飾。

先固定輸出比例

依實際刊登位置選擇正方形、橫向或直向比例。目前介面不提供生成後的擴圖或重新構圖功能，因此應在送出前確認比例。

畫面文字要寫清楚並複查

如果圖片需要出現文字，請明確提供文案、語言、位置與風格。

選擇它的理由

每次固定 1 點數，生成前即可確認成本。
操作精簡，只需提示詞與長寬比即可送出。
5 種比例涵蓋常見的正方形、橫向與直向構圖。
提示詞最多可輸入 1,000 個字元，足以組織清楚的場景描述。

使用前須知

目前只支援文字生成圖片，不接受參考圖，也不能編輯現有圖片。
提示詞上限為 1,000 個字元，冗長或限制很多的需求必須精簡。
介面未開放解析度、畫質、隨機種子或反向提示詞等細部控制。
畫面文字、細節與主體一致性仍需人工檢查；單價低不代表可以省略驗收。

可調整參數

協助你快速評估此模型是否符合你的業務情境。

長寬比
選填
參數類型: 下拉選單
預設值：1:1
1:1
4:3
3:4
16:9
9:16

點數消耗

公開資料 0.8／圖；本地 1 點數

公開資料服務目前原始價格為每張圖 0.8 credits；Kyeo 保留這個小數事實，並對單次總價執行 `Math.ceil`，因此單圖預授權與結算為 1 點數。

省點數小技巧

批次生成圖片或影片前，建議先用同一組素材，在目前模型與同類候選模型之間進行 A/B 測試；確認效果後再批次生成，避免浪費點數。

常見問題

同類模型推薦

還沒決定該用哪個嗎？順便比較一下這幾個同類候選模型。

Nano Banana

Nano Banana（Gemini 2.5 Flash Image）在 Kyeo AI 同時提供文字生成圖片與自然語言圖片編輯。未上傳圖片時使用 `Nano Banana`，上傳後則使用 `Nano Banana`；兩條路徑皆為 3 點數／次，提示詞最多 5,000 個字元。

AI 圖片生成與編輯模型
3 點數 / 次
查看詳情立即使用

GPT Image 1.5

GPT Image 1.5 在 Kyeo 支援文字生成圖片與圖生圖，可選 3 種長寬比與 medium／high 品質。圖生圖最多使用 16 張 JPG、PNG 或 WebP 參考圖，單張不超過 10 MB；medium 為 4 點數，high 為 22 點數。 GPT Image 1.5 是 OpenAI 已棄用的上一代圖片模型。Kyeo 仍保留其生成與編輯入口，但這不等於 OpenAI 仍將它列為目前首選。

AI 圖片模型
4-22 點數 / 次
查看詳情立即使用

Flux 2 Flex

Flux 2 Flex 在 Kyeo 提供文字生成圖片與最多 8 張圖片的編輯入口。未上傳圖片時使用 Flux 2 Flex，上傳後切換至 Flux 2 Flex；目前可選 7 種固定長寬比和 1K／2K，每次分別消耗 14／24 點數。

AI 圖片模型
14–24 點數／次
查看詳情立即使用

資訊來源

頁面內容依據廠商資料、模型系列說明與 Kyeo AI 平台串接設定整理。

來源說明

瞭解 Kyeo AI 目前開放的 Z-Image 文字生成圖片功能、1,000 個字元的提示詞上限、5 種長寬比，以及公開資料每張圖 0.8 credits、本地向上取整為 1 點數的計費方式。 公開資料服務目前原始價格為每張圖 0.8 credits；Kyeo 保留這個小數事實，並對單次總價執行 `Math.ceil`，因此單圖預授權與結算為 1 點數。

最後更新: 2026-08-04
Tongyi-MAI / Z-Image GitHub
官方
查看原始來源
Z-Image Technical Report (arXiv)
官方
查看原始來源